Unmanned Aerial Vehicles (UAVs) is a kind of aircraft performing certain intelligence without pilot. To fulfill complex and practical tasks, accurate state estimation is an essential subject of UAVs. However, UAV is easy to be lost and hardly to be localized again in the unknown environments due to low features and strong flexibility in the previous SLAM system. In this paper, a rectification strategy is proposed to improve UAVs' robustness of SLAM and autonomous navigation with monocular camera. The features in each frame are detected and compared with existed features in the world coordinate system. Then, we set a threshold value and construct a feedback system to make sure the UAV can always be tracked. Experimental result shows that the proposed method can make the SLAM algorithm more robust.